International audience

International audience

In this article, we propose an original solution to the problem of surface reconstruction of large-scale unknown environments, with multiple cooperative robots. As they progress through the 3-D environment, the robots rely on volumetric maps obtained via a TSDF representation to extract discrete incomplete surface elements (ISEs), and a list of candidate viewpoints is generated to cover them. A next-best-view planning approach, which approximately solves a traveling salesman problem (TSP) via greedy allocation, is then used to iteratively assign these viewpoints to the robots. Two multiagent architectures, a centralized one (TSP-Greedy Allocation or TSGA) and a distributed one (dist-TSGA), in which the robots locally compute their maps and share them, are developed and compared. Extensive numerical and real-world experiments with multiple aerial and ground robots in challenging 3-D environments show the flexibility and effectiveness of our surface representation of a volumetric map. The experiments also shed light on the nexus between reconstruction accuracy and surface completeness, and between total distance traveled and execution time.

International audience

This paper introduces an Online Localisation and Colored Mesh Reconstruction (OLCMR) ROS perception architecture for ground exploration robots aiming to perform robust Simultaneous Localisation And Mapping (SLAM) in challenging unknown environments and provide an associated colored 3D mesh representation in real time. It is intended to be used by a remote human operator to easily visualise the mapped environment during or after the mission or as a development base for further researches in the field of exploration robotics. The architecture is mainly composed of carefully-selected open-source ROS implementations of a LiDAR-based SLAM algorithm alongside a colored surface reconstruction procedure using a point cloud and RGB camera images projected into the 3D space. The overall performances are evaluated on the Newer College handheld LiDAR-Vision reference dataset and on two experimental trajectories gathered on board of representative wheeled robots in respectively urban and countryside outdoor environments.

Transferring learning-based models to the real world remains one of the hardest problems in model-free control theory. Due to the cost of data collection on a real robot and the limited sample efficiency of Deep Reinforcement Learning algorithms, models are usually trained in a simulator which theoretically provides an infinite amount of data. Despite offering unbounded trial and error runs, the reality gap between simulation and the physical world brings little guarantee about the policy behavior in real operation. Depending on the problem, expensive real fine-tuning and/or a complex domain randomization strategy may be required to produce a relevant policy. In this paper, a Soft-Actor Critic (SAC) training strategy using incremental environment complexity is proposed to drastically reduce the need for additional training in the real world. The application addressed is depth-based mapless navigation, where a mobile robot should reach a given waypoint in a cluttered environment with no prior mapping information. Experimental results in simulated and real environments are presented to assess quantitatively the efficiency of the proposed approach, which demonstrated a success rate twice higher than a naive strategy.

This article introduces a decentralized multi-robot algorithm for Simultaneous Localization And Mapping (SLAM) inspired from previous work on collaborative mapping [1]. This method makes robots jointly build and exchange i) a collection of 3D dense locally consistent submaps, based on a Truncated Signed Distance Field (TSDF) representation of the environment, and ii) a pose-graph representation which encodes the relative pose constraints between the TSDF submaps and the trajectory keyframes, derived from the odometry, inter-robot observations and loop closures. Such loop closures are spotted by aligning and fusing the TSDF submaps. The performances of this method have been evaluated on multi-robot scenarios built from the EuRoC dataset [2].

In this paper, we propose a novel cluster-based Next-Best-View path planning algorithm to simultaneously explore and inspect large-scale unknown environments with multiple Unmanned Aerial Vehicles (UAVs). In the majority of existing informative path-planning methods, a volumetric criterion is used for the exploration of unknown areas, and the presence of surfaces is only taken into account indirectly. Unfortunately, this approach may lead to inaccurate 3D models, with no guarantee of global surface coverage. To perform accurate 3D reconstructions and minimize runtime, we extend our previous online planner based on TSDF (Truncated Signed Distance Function) mapping, to a fleet of UAVs. Sensor configurations to be visited are directly extracted from the map and assigned greedily to the aerial vehicles, in order to maximize the global utility at the fleet level. The performances of the proposed TSGA (TSP-Greedy Allocation) planner and of a nearest neighbor planner have been compared via realistic numerical experiments in two challenging environments (a power plant and the Statue of Liberty) with up to five quadrotor UAVs equipped with stereo cameras.

The estimation of disparity maps from stereo pairs has many applications in robotics and autonomous driving. Stereo matching has first been solved using model-based approaches, with real-time considerations for some, but today's most recent works rely on deep convolutional neural networks and mainly focus on accuracy at the expense of computing time. In this paper, we present a new method for disparity maps estimation getting the best of both worlds: the accuracy of data-based methods and the speed of fast model-based ones. The proposed approach fuses prior disparity maps to estimate a refined version. The core of this fusion pipeline is a convolutional neural network that leverages dilated convolutions for fast context aggregation without spatial resolution loss. The resulting architecture is both very effective for the task of refining and fusing prior disparity maps and very light, allowing our fusion pipeline to produce disparity maps at rates up to 125 Hz. We obtain state-of-the-art results in terms of speed and accuracy on the KITTI benchmarks. Code and pre-trained models are available on our github: https://github.com/ferreram/FD-Fusion.

The characterization in real-time of the drivable space in front of the vehicle is a key issue for safe autonomous navigation or driving assistance. This paper presents a method that uses a lidar (a multilayer laser scanner) integrated in the front bumper of an automotive vehicle. A grid processing is first applied to detect and localize objects in the immediate environment after having compensated the movement of the vehicle. Accurate map information is then introduced in the perception scheme to refine the characterization of the drivable space. The paper details the different processing stages necessary to implement this method and presents the design of the system that has been prototyped on board an experimental vehicle. We report real experiments carried out in challenging urban environments to illustrate the performance of this approach which has been evaluated thanks to a precise retro-projection of the estimated drivable space in a wide-angle scene camera.

In urban environments, moving obstacles detection and free space determination are key issues for driving assistance systems and autonomous vehicles. When using lidar sensors scanning in front of the vehicle, uncertainty arises from ignorance and errors. Ignorance is due to the perception of new areas and errors come from imprecise pose estimation and noisy measurements. Complexity is also increased when the lidar provides multi-echo and multi-layer information. This paper presents an occupancy grid framework that has been designed to manage these different sources of uncertainty. A way to address this problem is to use grids projected onto the road surface in global and local frames. The global one generates the mapping and the local one is used to deal with moving objects. A credibilist approach is used to model the sensor information and to do a global fusion with the world-fixed map. Outdoor experimental results carried out with a precise positioning system show that such a perception strategy increases significantly the performance compared to a standard approach.

Advanced Driving Assistance Systems exploit exteroceptive sensors to help the driver in perceiving the dynamic environment, like other vehicles or pedestrians. This paper proposes an original approach to deal with this perception challenge in urban environments. The method detects mobile objects motions using grids elaborated thanks to a lidar range scanner and an enhanced map of the drivable space. The data fusion is performed using the Dempster-Shafer theory which provides an interesting framework particularly well adapted to manage the uncertainties of the sensors. By analyzing conflicting information, objects movements can be efficiently characterized. This formalism provides also the interesting possibility to introduce decay factors that are useful for forgetting old information. Experimental results obtained with an IBEO Alasca and an Applanix positioning system show that such a perception strategy can be effective compared to deterministic accumulation strategies.

In urban environments, detection of moving obstacles and free space determination are key issues for driving assistance systems or autonomous vehicles. This paper presents a lidar-based perception system for passenger-cars, able to do simultaneously mapping and moving obstacles detection. Nowadays, many lidars provide multi-layer and multi-echo measurements. A smart way to handle this multi-modality is to use grids projected on the road surface in both global and local frames. The global one generates the mapping and the local is used to deal with moving objects. An approach based on both positive and negative accumulation has been developed to address the remnant problem of quickly moving obstacles. This method is also well suited for multi-layer and multi-echo sensors. Experimental results carried out with an IBEO Alasca and an Applanix positioning system show the performance of such a perception strategy.